A fast matching algorithm with an adaptive window based on quasi-dense method

This paper presents a fast quasi-dense matching algorithm using an adaptive window. The algorithm starts from a set of sparse seed matches, and then propagates to the neighboring pixels; finally the most points in the images are matched. During computing the normalized cross correlation (NCC), an additional level of incremental calculation is proposed to achieve further speed-up. The confidence coefficient is introduced and the search window is varied in inverse proportion to it. The algorithm has been tested with stereo images and the results demonstrate its accuracy and efficiency. The algorithm also can be applied to a wide range of image pairs including those with large disparity or without rectification even if part of the images is less textured. In particular, with big images and a large disparity range our algorithm turns out to be significantly faster.